- developers - lists.mariadb.org

[Maria-developers] MariaDB config files order
by Honza Horak 20 Aug '13

20 Aug '13

Hi guys, maybe I just need some little kick or I should get a coffee, but I can't understand the following config file ordering difference. Reading [1] and code in mariadb-5.5.32/mysys/default.c:1226, it looks like expected order of config files in Linux should be: /etc/my.cnf /etc/mysql/my.cnf ~/.my.cnf However, when asking for that using --help --verbose, I get a bit different output: $ /usr/libexec/mysqld --help --verbose 2>/dev/null | grep --after=1 '^Default options' | tail -n 2 Default options are read from the following files in the given order: /etc/mysql/my.cnf /etc/my.cnf ~/.my.cnf I don't understand why /etc/my.cnf is not on the first place in that output. Can somebody give me a hint? Am I just blind? Just FTR I'm trying that on Fedora 19, but there are no changes in Fedora packages that would do anything related, so I guess the same result would be in other distros as well. [1] https://kb.askmonty.org/en/mysqld-startup-options/ Cheers, Honza

2 3

Re: [Maria-developers] [Commits] Rev 3677: Fix bug MDEV-4895 Valgrind warnings (Conditional jump or move depends on uninitialised value) in Field_datetime::get_date on GREATEST(..) IS NULL in file:///home/tsk/mprog/src/5.3/
by Alexander Barkov 20 Aug '13

20 Aug '13

Hi Timour, The patch looks Okey to push for me. Please find two small neat-peaking comments below though :) On 08/20/2013 04:36 PM, timour(a)askmonty.org wrote: > At file:///home/tsk/mprog/src/5.3/ > > ------------------------------------------------------------ > revno: 3677 > revision-id: timour(a)askmonty.org-20130820123610-cjtgvaybl1ip3128 > parent: igor(a)askmonty.org-20130815235920-io2h7tlypwlbunsp > fixes bug: https://mariadb.atlassian.net/browse/MDEV-4895 > committer: timour(a)askmonty.org > branch nick: 5.3 > timestamp: Tue 2013-08-20 15:36:10 +0300 > message: > Fix bug MDEV-4895 Valgrind warnings (Conditional jump or move depends on uninitialised value) in Field_datetime::get_date on GREATEST(..) IS NULL > > Analysis: > The cause of the valgrind warning was an attempt to evaluate a Field that was not yet read. > The reason was that on one hand Item_func_isnotnull was marked as constant by > Item_func_isnotnull::update_used_tables, and this allowed eval_const_cond() to be called. > On the other hand Item_func_isnotnull::val_int() evaluated its argument as if it was not > constant. > > Solution: > The fix make sure that Item_func_isnotnull::val_int() doesn't evaluate its argument when > it is constant and cannot be NULL, because the result is known in this case. > > > > === modified file 'mysql-test/r/null.result' > --- a/mysql-test/r/null.result 2011-11-21 13:16:16 +0000 > +++ b/mysql-test/r/null.result 2013-08-20 12:36:10 +0000 > @@ -343,3 +343,33 @@ Field Type Null Key Default Extra > IFNULL(NULL, b) decimal(1,0) YES NULL > DROP TABLE t1, t2; > # End of 5.0 tests > +# > +# MDEV-4895 Valgrind warnings (Conditional jump or move depends on uninitialised value) in Field_datetime::get_date on GREATEST(..) IS NULL > +# > +CREATE TABLE t1 (dt DATETIME NOT NULL); > +INSERT INTO t1 VALUES (NOW()),(NOW()); > +EXPLAIN > +SELECT * FROM t1 WHERE concat( dt, '2012-12-21 12:12:12' ) IS NULL; > +id select_type table type possible_keys key key_len ref rows Extra > +1 SIMPLE NULL NULL NULL NULL NULL NULL NULL Impossible WHERE > +SELECT * FROM t1 WHERE concat( dt, '2012-12-21 12:12:12' ) IS NULL; > +dt > +drop table t1; > +CREATE TABLE t1 (dt INT NOT NULL); > +INSERT INTO t1 VALUES (1),(2); > +EXPLAIN > +SELECT * FROM t1 WHERE concat( dt, '1' ) IS NULL; > +id select_type table type possible_keys key key_len ref rows Extra > +1 SIMPLE NULL NULL NULL NULL NULL NULL NULL Impossible WHERE > +SELECT * FROM t1 WHERE concat( dt, '1' ) IS NULL; > +dt > +drop table t1; > +CREATE TABLE t1 (dt INT NOT NULL); > +INSERT INTO t1 VALUES (1),(2); > +EXPLAIN > +SELECT * FROM t1 WHERE NOT (concat( dt, '1' ) IS NOT NULL); > +id select_type table type possible_keys key key_len ref rows Extra > +1 SIMPLE NULL NULL NULL NULL NULL NULL NULL Impossible WHERE > +SELECT * FROM t1 WHERE NOT (concat( dt, '1' ) IS NOT NULL); > +dt > +drop table t1; Please fix "drop table" to upper case here, and in the other places. > > === modified file 'mysql-test/t/null.test' > --- a/mysql-test/t/null.test 2009-02-10 03:00:11 +0000 > +++ b/mysql-test/t/null.test 2013-08-20 12:36:10 +0000 > @@ -255,3 +255,31 @@ DESCRIBE t2; > DROP TABLE t1, t2; > > --echo # End of 5.0 tests > + > +--echo # > +--echo # MDEV-4895 Valgrind warnings (Conditional jump or move depends on uninitialised value) in Field_datetime::get_date on GREATEST(..) IS NULL > +--echo # > + > +CREATE TABLE t1 (dt DATETIME NOT NULL); > +INSERT INTO t1 VALUES (NOW()),(NOW()); > + > +EXPLAIN > +SELECT * FROM t1 WHERE concat( dt, '2012-12-21 12:12:12' ) IS NULL; > +SELECT * FROM t1 WHERE concat( dt, '2012-12-21 12:12:12' ) IS NULL; > + > +drop table t1; > +CREATE TABLE t1 (dt INT NOT NULL); > +INSERT INTO t1 VALUES (1),(2); > +EXPLAIN > +SELECT * FROM t1 WHERE concat( dt, '1' ) IS NULL; > +SELECT * FROM t1 WHERE concat( dt, '1' ) IS NULL; > + > +drop table t1; > +CREATE TABLE t1 (dt INT NOT NULL); > +INSERT INTO t1 VALUES (1),(2); > + > +EXPLAIN > +SELECT * FROM t1 WHERE NOT (concat( dt, '1' ) IS NOT NULL); > +SELECT * FROM t1 WHERE NOT (concat( dt, '1' ) IS NOT NULL); > + > +drop table t1; > > === modified file 'sql/item_cmpfunc.cc' > --- a/sql/item_cmpfunc.cc 2013-08-15 23:59:20 +0000 > +++ b/sql/item_cmpfunc.cc 2013-08-20 12:36:10 +0000 > @@ -4621,6 +4621,8 @@ Item *and_expressions(Item *a, Item *b, > longlong Item_func_isnull::val_int() > { > DBUG_ASSERT(fixed == 1); > + if (const_item() && !args[0]->maybe_null) > + return false; "return 0" should be better in a longlong function. > return args[0]->is_null() ? 1: 0; > } > > @@ -4628,6 +4630,8 @@ longlong Item_is_not_null_test::val_int( > { > DBUG_ASSERT(fixed == 1); > DBUG_ENTER("Item_is_not_null_test::val_int"); > + if (const_item() && !args[0]->maybe_null) > + DBUG_RETURN(1); > if (args[0]->is_null()) > { > DBUG_PRINT("info", ("null")); >

1 0

Re: [Maria-developers] [Commits] MDEV-4923 file://home/jan/mysql/mariadb-5.5-mdev4923
by Sergei Golubchik 20 Aug '13

20 Aug '13

Hi, Jan! On Aug 20, Jan Lindström wrote: > Serg, > > Could you please review fix for MDEC-4923, where these changes are > pushed lp:maria/5.5 or lp:~maria-captains/maria/5.5 ? Looks correct. This seems to be XtraDB bug. lp:maria/5.5 is an alias for lp:~maria-captains/maria/5.5 > === modified file 'storage/xtradb/os/os0file.c' > --- storage/xtradb/os/os0file.c 2013-07-29 14:03:41 +0000 > +++ storage/xtradb/os/os0file.c 2013-08-20 05:35:22 +0000 > @@ -1658,8 +1658,8 @@ > } > } > > - if (srv_use_atomic_writes && type == OS_DATA_FILE && > - os_file_set_atomic_writes(file, name)) { > + if (srv_use_atomic_writes && type == OS_DATA_FILE && > + !os_file_set_atomic_writes(file, name)) { > CloseHandle(file); > *success = FALSE; > file = INVALID_HANDLE_VALUE; > @@ -1786,7 +1786,7 @@ > #endif /* USE_FILE_LOCK */ > > if (srv_use_atomic_writes && type == OS_DATA_FILE > - && os_file_set_atomic_writes(name, file)) { > + && !os_file_set_atomic_writes(name, file)) { > > *success = FALSE; > close(file); > Regards, Sergei

1 0

[Maria-developers] [Otto Kekäläinen] Fwd: initial feedback on mariadb
by Kristian Nielsen 20 Aug '13

20 Aug '13

Hi, Otto asked for help with the copyright file for the Debian MariaDB packages. As I understand it, the .deb packages have a file /usr/share/doc/mariadb-server-5.5/copyright and there are some scripts in Debian that check the file against what is in the headers of source files and possibly other checks (error output attached below). Otto did something, but there is apparently more needed. Can some of you who understand the finer points of licensing among Oracle, Monty Program, MariaDB Services, MariaDB foundation, Percona, ... please look into this and help Otto fix up things? Debian are usually very careful with licensing, so this could easily become a blocker otherwise (I assume). - Kristian.

4 4

[Maria-developers] MDEV-4911 - KILL command with query id
by Roberto Spadim 20 Aug '13

20 Aug '13

Hi guys, i sent some patchs, could anyone check it? it have some things to do: query syntax isn't in the MDEV-4911, i did a QUERY_ID instead of a more comlpex syntax (since i never used bison before :P ) the query_id works, but i need a new error message ( "Unknown query id: %") well i didn't make the gcc happy, there's some warnings, and i don't know if it's all right i send this patch after MDEV-4916 patch, in other words the %expect value of sql_yacc.yy is wrong if you don't apply the MDEV-4916 patch the 4916 is a IF_IDLE keyword to KILL command too it only kill if there's no query running well :) thanks guys! <https://mariadb.atlassian.net/browse/MDEV-4916> -- Roberto Spadim

2 3

Re: [Maria-developers] More Secure Kill Command
by Sergei Golubchik 20 Aug '13

20 Aug '13

Hi, Roberto! On Aug 17, Roberto Spadim wrote: > Hi sergei, > What about give two informations, thread id and query id > If thread id and query id is equal, kill it There's no need to. Query id identifies the query and the connection unambiguously. Specifying also a thread id is redundant. Regards, Sergei

2 9

[Maria-developers] linking an external library
by rg 19 Aug '13

19 Aug '13

Folks, I want to link an external library for a language extension. After several hours of trying to tell cmake (^%$@*&) to pick it up for linking the library would still not be built with. CMakeLists.txt in trunk/sql is in the meantime full of pointers to the library without avail. I did the normal FIND_LIBRARY() followed by TARGET_LINK_LIBRARIES({$LIBNAME}) on various targets plus as nothing worked: SET(CMAKE_LIBRARY_PATH) on it, explicity LINK_DIRECTORIES(), SET(CMAKE_PREFIX_PATH), SET(CMAKE_LIBRARY_PATH) ... How do I link mariadb with an external library?? Thanks Rob

3 4

[Maria-developers] Audit plugin and qc_info plugin - MDEV-4682
by Roberto Spadim 18 Aug '13

18 Aug '13

Hi guys, i put some fields in qc_info plugin https://mariadb.atlassian.net/browse/MDEV-4581 the second part was add some others informations (stats): https://mariadb.atlassian.net/browse/MDEV-4682 i was reading the "response_time_distribution" of percona, patch at https://mariadb.atlassian.net/browse/MDEV-4568 could i use the audit plugin to implement MDEV-4682? what i need is: 1)know if the query was executed with/without query cache 2)know the time expend (lock time, and total time) to execute the query (first execution) 3)know the time expend with query cache hit 4) both (3) and (2) don't include time of network packages, just server side times... is it possible? any idea how could i implement it? i didn't write a audit plugin yet =( well.. time to learn =D thanks guys! -- Roberto Spadim

2 2

[Maria-developers] ALTER ONLINE TABLE syntax
by Sergei Golubchik 18 Aug '13

18 Aug '13

Hi, You've added this ALTER ONLINE TABLE syntax, where ONLINE de facto means "without copying the data". Now, after 5.6 merge we also have ALGORITHM=COPY|INPLACE and LOCK=NONE|SHARED|EXCLUSIVE. I've mapped ONLINE to ALGORITHM=INPLACE, because that's what it means. But it's confusing, one can have an "inplace" operation under an exclusive lock, if the engine wants it that way. And it won't be "online" by any stretch of imagination. One can also technically have a "copy" operation with a very permissive write-allow-write lock, and it'll be as online as one can get it. I mean, logically ONLINE should mean LOCK=NONE, not ALGORITHM=INPLACE. But that would change the existing behavior and the test results too. For example, *any* ALTER on the temporary table would always be "online", because, well, it wouldn't take any lock on that table. Alternatively, we can keep ONLINE as ALGORITHM=INPLACE, and deprecate it, issue a warning, suggesting to use the new 5.6 syntax. Opinions? Regards, Sergei

5 7

[Maria-developers] MDEV-4916
by Roberto Spadim 18 Aug '13

18 Aug '13

Patch uploaded, and test work (but i didn't add a unit test yet :/ sorry i don't know yet how to write it) -- Roberto Spadim

1 0